基于脑电运动速度想象的单次识别研究

The Study of Single-trial Identification of Imagined Movement Speeds Based on EEG

基于运动想象脑电节律活动的脑-机接口是脑-机接口系统研究中的重要范式之一。本研究给出一种基于运动速度想象的新的研究范式,探索在该研究范式下对运动速度想象具有反应性的脑电节律活动,并进行单次识别。采集了4个健康志愿者想象左手食指快速运动(4 Hz)和慢速运动(1 Hz)时的脑电信号,速度由节拍器定节奏和训练。通过能量谱分析,在C3、Cz和C4通道发现了对运动速度想象具有反应性的频带:9 Hz至13 Hz。提取通道C3、Cz和C4上9 Hz至13 Hz频带能量构建特征空间,分别利用Fisher判别分析和多层感知器神经网络进行运动速度想象的单次识别,对于左手食指快速运动和慢速运动想象,Fisher判别分析和多层感知器神经网络取得的平均误分类率分别是27.7±1.2%,28.4±4.6%,正确识别率均在70%以上。结果表明,尽管运动速度想象的单次识别是一个困难的挑战,但通过精心设计研究范式,适当训练被试,能够诱发出对速度起反应的特征频带,基于脑电单次识别运动速度想象是可行的,该研究可望能够为脑-机接口提供额外的新的速度控制参数。

Brain-computer interface based on EEG rhythmic activities evoked by motor imagery is one of the important paradigms for brain-computer interface systems.A new paradigm based on imagined movement speeds was presented in the study.EEG rhythmic activities reactive to imagined movement speeds were explored and identification of single-trial EEG related to imagined movement speeds was investigated under the new paradigm.EEG signals were acquired from 4 healthy subjects during imagining their left index fingers movement at two speeds（4 Hz and 1 Hz,trained and paced by metronome）.A rhythmic frequency band around 9~13 Hz over C3,Cz,and C4 related to imagined movement speeds was detected by spectrum analysis.Feature space was built from the band power of 9~13 Hz at C3,CZ,and C4.Fisher discriminant analysis（FDA） and multi-layer perception neural network（MLP） were applied in single-trial identification of imagined movement speeds respectively.The averaged misclassification rates between fast and slow movement imagination involved in left index fingers with FDA and MLP were 27.7±1.2% and 28.4±4.6% respectively.The accurate recognition rates were above 70%.The results show that although single-trial identification for imagined movement speeds based on EEG is a challenging research,distinctive frequency band activities related to imagined speeds can be evoked by carefully designing research paradigms and properly training subjects,and single-trial identification of imagined movement speeds based on EEG is feasible.The study is expected to provide additional new speed control parameters for brain-computer interface.